Document 10940114

advertisement
Hindawi Publishing Corporation
Journal of Inequalities and Applications
Volume 2009, Article ID 409809, 20 pages
doi:10.1155/2009/409809
Research Article
Generalized Bihari Type Integral Inequalities and
the Corresponding Integral Equations
László Horváth
Department of Mathematics, University of Pannonia, Egyetem u. 10, 8200 Veszprém, Hungary
Correspondence should be addressed to László Horváth, lhorvath@almos.vein.hu
Received 2 February 2009; Accepted 23 June 2009
Recommended by Alberto Cabada
We study some special nonlinear integral inequalities and the corresponding integral equations
in measure spaces. They are significant generalizations of Bihari type integral inequalities
and Volterra and Fredholm type integral equations. The kernels of the integral operators are
determined by concave functions. Explicit upper bounds are given for the solutions of the integral
inequalities. The integral equations are investigated with regard to the existence of a minimal and
a maximal solution, extension of the solutions, and the generation of the solutions by successive
approximations.
Copyright q 2009 László Horváth. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.
1. Introduction and the Main Results
In this paper we study integral inequalities of the form
yx ≤ fx gx
q ◦ y dμ,
x ∈ X,
1.1
q ◦ y dμ,
x ∈ X,
1.2
Sx
and the corresponding integral equations
yx fx gx
Sx
where
A1 X, A, μ is a measure space;
A2 S is a function from X into A such that the following properties hold:
2
Journal of Inequalities and Applications
A12 μSx < ∞ for every x ∈ X,
A22 if x2 ∈ Sx1 , then Sx2 ⊂ Sx1 ,
A32 {x1 , x2 ∈ X 2 | x2 ∈ Sx1 } is μ2 -measurable;
A3 q is a function from 0, ∞ into 0, ∞ with the following conditions:
A13 q is concave,
A23 limt → ∞ qt/t 0;
A4 the functions f and g belong to
Lloc X : p : X −→ 0, ∞ | p is μ-integrable over Sx ∀x ∈ X .
1.3
It may be noted that under the condition A13 the function q is increasing see Lemma 2.1 for
the justification.
A always represents a σ-algebra in X. The μ-integrable functions over a measurable
set A ∈ A are considered to be μ-almost measurable on A. The product of the measure space
X, A, μ with itself is understood as in 1, and it is denoted by X 2 , A2 , μ2 .
By N we designate the set of nonnegative integers.
Special cases of 1.1 seem first to have been investigated by Lasalle 2 and Bihari
3. Bihari’s classical result gives an explicit upper bound for the solutions of the integral
inequality
ut ≤ a t
kshusds,
t ∈ c, d,
1.4
c
where a ≥ 0, k and u are nonnegative continuous functions on c, d, and h is a positive
continuous and increasing function on 0, ∞. A group of inequalities is now associated with
Bihari’s name. Results for the various forms of such inequalities and references to different
works in this topic can be found in 4–6. Bihari type inequalities have been widely studied
because they can be applied in the theory of difference, differential and integral equations.
Riemann or classical Lebesgue integral is used in most of the theorems in this area. There are
relatively few papers using other types of integral. For generalizations to abstract Lebesgue
integral; see 7–9. The linear version of 1.1 is given in 7. The special case qx xα ,
x ≥ 0, 0 < α < 1 of 1.1 is considered in 8, while the special case qx xα , x ≥ 0,
1 < α of 1.1 is discussed in 9. It turns out to be useful to study Bihari type inequalities
with abstract Lebesgue integral. It is motivated proceeding in this direction as follows. We
can get new facts about the nature of Bihari type inequalities even in the finite dimensional
environment; the results can be applied in the study of certain new classes of differential and
integral equations see 7–11.
The traditional treatment assumes not only that X ⊂ Rn , but also that the sets Sx,
x ∈ X are intervals, while the present treatment it should be emphasized that the methods
employed to establish our results are not usual in this topic makes it possible to consider
more general sets examples for functions satisfying A22 and A32 can be found in 11.
Such results are not quite so easy to find in literature, although they can be used as powerful
tools in many fields of mathematics.
Besides Lloc X, the following function spaces will play an important role.
Journal of Inequalities and Applications
3
Definition 1.1. If A is a nonempty subset of X such that Sx ⊂ A for every x ∈ A, then let
Lloc A : p : A −→ 0, ∞ | p is μ-integrable over Sx ∀x ∈ A .
1.5
Next, the basic concepts of the solutions of the inequalities 1.1 and
yx ≥ fx gx
q ◦ y dμ,
x ∈ X,
1.6
Sx
and the equation 1.2 are defined.
Definition 1.2. We say that a function y : Dy → R is a solution of 1.1, 1.6, or 1.2 if
i Dy is a nonempty subset of X such that Sx ⊂ Dy for every x ∈ Dy ,
ii y ∈ Lloc Dy ,
iii y satisfies 1.1, 1.6, or 1.2 for each x ∈ Dy .
It is easily verified see Lemma 2.5 that if y : Dy → R is a solution of 1.1, 1.6, or
1.2, then q ◦ y is μ-integrable over Sx for all x ∈ Dy .
After these preparations we set ourselves the task of obtaining an upper bound for the
solutions of 1.1. The following definition will be useful.
Definition 1.3. a For every x ∈ X with μSx > 0, let
1
ax :
μSx
f dμ,
bx :
Sx
1.7
g dμ.
Sx
b Let
tx :
⎧
⎨max t ≥ 0 | t ax bx qt ,
if μSx > 0,
⎩0,
if μSx 0,
x ∈ X.
1.8
By A23 , tx is a nonnegative real number for every x ∈ X.
Now we are in a position to formulate the first main result.
Theorem 1.4. Assume the conditions (A1 )–(A4 ).
a Every solution y : Dy → R of 1.1 satisfies
yx ≤ fx gxμSxqtx,
x ∈ Dy .
1.9
b The function z defined on X by
zx fx gxμSxqtx
belongs to Lloc X.
1.10
4
Journal of Inequalities and Applications
In the second main result we test the scope of the previous theorem by applying it to
prove the existence of a maximal and a minimal solution of the integral equation 1.2. At the
same time, we show that every solution has maximal domain of existence X, and we apply
the method of successive approximations to 1.2. Moreover, the behavior of the solutions is
studied in a special case. The considered integral equations are in a very general form, there
are classical Volterra and Fredholm type integral equations among them.
Theorem 1.5. Suppose the conditions (A1 )–(A4 ).
a1 There exists a solution ymin ∈ Lloc X of 1.2, which is minimal in the sense that
ymin x ≤ yx, x ∈ Dy whenever y : Dy → R is a solution of 1.6.
a2 There exists a solution ymax ∈ Lloc X of 1.2, which is maximal in the sense that
ymax x ≥ yx, x ∈ Dy whenever y : Dy → R is a solution of 1.1.
b If y : Dy → R is a solution of 1.2, then y has an extension y to X that is a solution of
1.2 on X.
c Let y : Dy → R be a solution of 1.1. Then the successive approximations determined by
y,
y0 : y,
q ◦ yn dμ,
yn1 x : fx gx
x ∈ Dy , n ∈ N,
1.11
Sx
are well defined, yn ∈ Lloc Dy , n ∈ N; the sequence yn ∞
n0 is increasing and converges
pointwise on Dy to a solution of 1.2.
d Let y : Dy → R be a solution of 1.6. Then the successive approximations 1.11
determined by y are well defined, yn ∈ Lloc Dy , n ∈ N, and the sequence yn ∞
n0 is
decreasing. Moreover, if either q is continuous (at 0) or q0 > 0, then they converge
pointwise on Dy to a solution of 1.2.
e If in addition f and g are bounded on Sx for all x ∈ X, then every solution y : Dy → R
of 1.2 is bounded on Sx for all x ∈ Dy .
We conclude this section with some remarks.
Remark 1.6. The next example shows that the concavity of q alone does not imply neither the
existence of an upper bound for the solutions of the integral inequality 1.1 nor the existence
of a solution of the integral equation 1.2. Consider the integral inequality
yx ≤ 1 0,x
q ◦ ydε0 1 q y0 ,
x ∈ 0, ∞,
1.12
x ∈ 0, ∞,
1.13
and the corresponding integral equation
yx 1 0,x
q ◦ ydε0 1 q y0 ,
where ε0 is the unit mass at 0 defined on the σ-algebra of Borel subsets of 0, ∞, and
q : 0, ∞ −→ 0, ∞,
qt : t.
1.14
Journal of Inequalities and Applications
5
Then the conditions A1 –A4 are satisfied without A23 . It is obvious that the functions
yn : 0, ∞ −→ 0, ∞,
yn x : n,
n∈N
1.15
are solutions of 1.12, showing that there are no either global or local upper bounds for the
solutions of 1.12. It is easy to check that 1.13 has no solution.
Remark 1.7. It is illustrated by an example that under the conditions A1 –A4 the maximal
domain of existence of a solution of 1.1 may be a proper subset of X. Let X : 0, 1, let A be
the Lebesgue measurable subsets of X, and let μ be the Lebesgue measure on A. The function
S is defined on A by
Sx :
⎧
⎨∅,
if x ∈ 0, 1,
⎩0, 1,
if x 1.
1.16
The functions f and g are defined on X by fx gx : 1. Suppose q : 0, ∞ → R, qt :
t1/2 . Then A1 –A4 are satisfied. Let y : 0, 1 → 0, 1 be a non Lebesgue measurable
function. Then y is a solution of 1.1 which has no extension to X.
Remark 1.8. The following example makes it clear that some extra conditions for q are
necessary in Theorem 1.5d. Let
X :
1
1 − | k ∈ N \ {0} ∪ {1, 2},
k
1.17
let A be the power set of X, and let the measure μ be defined on A by
μ :
∞
1
ε
ε1 ε2 ,
k 1−1/k
k1 2
1.18
where the measure εx x ∈ X is the unit mass at x defined on A. We consider the integral
equation
q ◦ y dμ,
yx x ∈ X,
1.19
Sx
where S : X → A, Sx : {s ∈ X | s < x}, and
q : 0, ∞ −→ 0, ∞,
qt :
⎧
⎨0,
if t 0,
⎩1,
if t > 0.
1.20
6
Journal of Inequalities and Applications
The conditions A1 –A4 can be immediately verified. Define the function y0 : X → R by
y0 x : 4. Noting that μX 3 leads to the inequality
q ◦ y0 dμ,
y0 x ≥
x ∈ X.
1.21
Sx
A few easy calculations imply that, for every n ∈ N \ {0},
⎧
0,
⎪
⎪
⎨
1
k−1
yn 1 −
1
⎪
k
⎪
,
⎩
i
2
in
yn 1 ∞
1
,
i
in 2
if 1 ≤ k ≤ n,
if k ≥ n 1,
yn 2 1 1.22
∞
1
in
2i
,
and therefore
lim yn x n→∞
⎧
⎨0,
⎩
1,
if x 1 −
1
, k ∈ N \ {0} or x 1,
k
if x 2.
1.23
Since 1.19 has the unique solution yx 0, x ∈ X, then the successive approximations
yn ∞
n0 do not converge to the solution of 1.19.
2. Preliminaries
This section is devoted to some preparatory results. In the following three lemmas we
establish some useful properties of concave functions.
Lemma 2.1. If the function r : 0, ∞ → 0, ∞ is concave, then r is increasing.
Proof. Suppose that there exist 0 ≤ t1 < t2 for which rt1 > rt2 . By the concavity of r, the
points of the graph of r are below or on the ray from t1 through t2 for all t ≥ t2 , and therefore
rt ≤
rt2 − rt1 t − t2 rt2 ,
t2 − t1
t ≥ t2 .
2.1
Since
rt2 − rt1 < 0,
t2 − t1
it follows from 2.1 that rt < 0 if t is large enough. This contradicts the range of r.
2.2
Journal of Inequalities and Applications
7
Lemma 2.2. Suppose the function r : 0, ∞ → R is concave, r0 ≥ 0, and limt → ∞ rt −∞.
Associate to r the nonnegative real number
tr : max{t ≥ 0 | rt 0}.
2.3
Then
a r is strictly decreasing on tr , ∞;
b rt ≥ 0 for all t ∈ 0, tr ;
c If tr > 0, and there is a t1 ∈ 0, tr such that rt1 0, then rt 0 for all t ∈ 0, tr .
Proof. The hypotheses on r since r is concave, r is continuous on 0, ∞ guarantee that
exactly one of the following three cases holds:
i r0 0 and rt < 0 for all t ∈ 0, ∞;
ii there exists a t > 0 such that rt 0 for all t ∈ 0, t and rt < 0 for all t ∈ t, ∞;
iii there is a unique t > 0 such that rt 0, rt > 0 for all t ∈ 0, t and rt < 0 for all
t ∈ t, ∞.
It follows that tr 0 if i is satisfied, and tr t otherwise. At the same time b and c are
proved.
It remains to show a. If tr < t, then i–iii show that 0 rtr > rt. Assume
tr < t1 < t2 . By the concavity of r,
rt1 rt2 ≥
,
t1 − tr t2 − tr
2.4
and hence
rt1 ≥
t1 − tr
rt2 > rt2 .
t2 − tr
2.5
The proof is complete.
Lemma 2.3. Suppose the function r : 0, ∞ → 0, ∞ is concave, and limt → ∞ rt/t 0.
Associate to r and to each of the nonnegative real numbers a, b the function
ra,b : 0, ∞ −→ R,
ra,b t : a brt − t.
2.6
Then
a ra,b is concave, and limt → ∞ ra,b t −∞;
b If t ≤ a brt, t ≥ 0, then t ∈ 0, ta,b , where
ta,b : max{t ≥ 0 | ra,b t 0};
2.7
8
Journal of Inequalities and Applications
c If a0 , b0 ∈ 0, ∞×0, ∞ and t0 > 0, then ra,b → ra0 ,b0 uniformly on 0, t0 as a, b →
a0 , b0 in 0, ∞×0, ∞;
d the function a, b → ta,b defined on 0, ∞×0, ∞ is upper semicontinuous.
Proof. a It is obvious.
b By a, Lemma 2.2b and a give the result.
c The triangle inequality insures that
|ra,b t − ra0 ,b0 t| ≤ |a − a0 | |b − b0 |rt,
t ∈ 0, t0 .
2.8
t ∈ 0, t0 ,
2.9
Then from Lemma 2.1
|ra,b t − ra0 ,b0 t| ≤ |a − a0 | |b − b0 |rt0 ,
and this gives the result.
d To prove this, choose a0 , b0 ∈ 0, ∞×0, ∞, and > 0. The definition of ta0 ,b0 and
Lemma 2.2a imply that
δ : −ra0 ,b0 ta0 ,b0 > 0.
2.10
By b, ra,b → ra0 ,b0 uniformly on 0, ta0 ,b0 as a, b → a0 , b0 in 0, ∞×0, ∞, and hence
there exists a neighborhood U of a0 , b0 in 0, ∞×0, ∞ such that
a, b ∈ U, t ∈ 0, ta0 ,b0 .
|ra,b t − ra0 ,b0 t| < δ,
2.11
It now follows from Lemma 2.2 a that
ta,b ≤ ta0 ,b0 ,
a, b ∈ U,
2.12
and the proof is complete.
The next result was proved in 8, Lemma 5b.
Lemma 2.4. Suppose that (A1 ) and (A32 ) hold. Let A ∈ A such that Sx ⊂ A for every x ∈ A.
Suppose u : A → R is μ-integrable over A, v : A → R is μ-almost measurable on A, and there
exists a measurable subset C of A such that μC is σ-finite and vx 0 for all x ∈ A \ C. Then the
function
x −→ vx
u dμ,
Sx
is μ-almost measurable on A.
x∈A
2.13
Journal of Inequalities and Applications
9
Lemma 2.5. Assume the conditions (A1 )–(A3 ). If A is a nonempty subset of X such that Sx ⊂ A
for every x ∈ A and u ∈ Lloc A, then q ◦ u ∈ Lloc A.
Proof. Let x ∈ A be fixed. Since q is increasing it is Borel measurable. Consequently, since u
is μ-almost measurable on Sx, q ◦ u is μ-almost measurable on Sx. By A23 , we can find
t0 > 0 such that qt ≤ t for all t ≥ t0 . Hence, note that q is increasing:
qus ≤
⎧
⎨us,
if us ≥ t0 ,
⎩qt ,
0
if us < t0 ,
s ∈ Sx.
2.14
It now follows from the definition of Lloc A and from A12 that q ◦ u is μ-integrable over
Sx. The proof is complete.
A consequence of the previous results that will be important later on is follows.
Lemma 2.6. Suppose that (A1 )–(A4 ) hold. If A is a nonempty subset of X such that Sx ⊂ A for
every x ∈ A and u ∈ Lloc A, then the function
q ◦ u dμ,
x −→ fx gx
x∈A
2.15
Sx
belongs to Lloc A.
Proof. Let x ∈ X.
By Lemma 2.4, the function
q ◦ u dμ,
s −→ gs
s∈A
2.16
Ss
is μ-almost measurable on Sx. Hence it follows from the μ-integrability of g and q ◦ u over
Sx the latter can be seen from Lemma 2.5, combined with the inequality
q ◦ u dμ ≤ gs
gs
Ss
q ◦ u dμ,
s ∈ Sx,
2.17
Sx
that the function 2.16 is μ-integrable over Sx. We conclude that the function 2.15 is μintegrable over Sx.
The proof is complete.
We need the concept of AL-space, which is of fundamental significance in the proof of
Theorem 1.5.
Definition 2.7. Suppose A1 , and let A be a nonempty set from A.
a Let
LA : u : A −→ R | u is μ-integrable over A .
For a given u ∈ LA, the symbol u is defined by u :
A
|u| dμ.
2.18
10
Journal of Inequalities and Applications
b Let N : {u ∈ LA | u 0}, and let LA : LA/N. For every u ∈ LA, let
u ∈ LA be the equivalence class containing u u u N, and we set u : u.
c We introduce the canonical ordering on LA: for u1 , u2 ∈ LA, and u1 u2 means
that u1 ≤ u2 μ-almost everywhere on A.
Remark 2.8. a LA, · is a complete pseudometric space.
b LA, · , is an L-normed Banach lattice, briefly, AL-space see 12.
If u is a function and A is a subset of the domain of u, then the restriction of u to A is
denoted by u | A.
Lemma 2.9. Suppose (A1 ), and let A and B be nonempty sets from A such that B ⊂ A. If F : {uλ |
λ ∈ Λ} and G : {vλ | λ ∈ Λ} are nonempty majorized subsets of LA such that uλ | B vλ |
B λ ∈ Λ, then
sup F | B sup G | B.
2.19
Proof. Since LA, · , is an AL-space, it is order complete see 12, and hence sup F and
sup G exist. Let u ∈ sup F and v ∈ sup G. Then u ≥ uλ μ-almost everywhere on A λ ∈ Λ,
thus the function
v1 ∈ LA,
v1 x :
⎧
⎨ux,
if x ∈ B,
⎩vx,
if x ∈ A \ B
2.20
is an upper bound of G. It follows that sup G v1 , that is v ≤ u μ-almost everywhere on B.
An argument entirely similar to the preceding part gives that sup F u1 , where
u1 ∈ LA,
u1 x :
⎧
⎨vx,
if x ∈ B,
⎩ux,
if x ∈ A \ B,
2.21
and therefore u ≤ v μ-almost everywhere on B.
The proof is now complete.
The next result can be found in 11, Lemma 16.
Lemma 2.10. Assume that the hypotheses (A1 ), (A22 ), (A32 ), and (A4 ) are satisfied. Let L : {x ∈ X |
Sx /
∅}. Suppose we are given solutions sx ∈ LSx, x ∈ L of 1.2 such that sx2 | Sx1 sx1
for each x1 ∈ L, x2 ∈ X with x1 ∈ Sx2 . Then there exists exactly one solution s : X → R of 1.2
for which s | Sx sx , x ∈ L.
Journal of Inequalities and Applications
11
3. Proofs of the Main Results
Consider now the proof of Theorem 1.4.
Proof. a If x ∈ Dy such that μSx 0, then 1.9 follows directly from 1.1.
Now, fix a point x ∈ Dy with μSx > 0. To estimate the second term on the right of
1.1, we can apply Jensen’s inequality see 13, by A12 :
ys ≤ fs gs
q ◦ y dμ
Ss
≤ fs gs
q ◦ y dμ
Sx
≤ fs gsμSxq
3.1
1
μSx
y dμ ,
s ∈ Sx,
Sx
and therefore
1
μSx
1
y dμ ≤
μSx
Sx
f dμ g dμ · q
Sx
Sx
1
μSx
y dμ .
3.2
Sx
This inequality, together with Definition 1.3a, implies that the expression
1
μSx
3.3
y dμ
Sx
satisfies the inequality
t ≤ ax bx qt,
t ≥ 0.
3.4
By Lemma 2.3 b and Definition 1.3 b, t ∈ 0, tx, and hence 1.9 can be deduced from
q ◦ y dμ ≤ fx gxμSxq
yx ≤ fx gx
Sx
1
μSx
y dμ .
3.5
Sx
b The properties defining Lloc X are trivial if x ∈ X with Sx g dμ 0. So assume
x ∈ X such that Sx g dμ > 0.
First, we show that the function z is μ-almost measurable on Sx. It is an easy
consequence of A12 and Lemma 2.4 that the functions
s −→ μSs 1 dμ,
Ss
s −→
f dμ,
Ss
s ∈ X,
3.6
s −→
g dμ,
Ss
s∈X
3.7
12
Journal of Inequalities and Applications
are μ-almost measurable on Sx. This means that there exists a measurable subset C of Sx
such that μSx \ C 0 and 3.6, 3.7 are measurable on C. Further, since g is μ-almost
measurable on Sx, it can be supposed that g is measurable on C. Thus we need to show
that the function
s −→ qts,
s∈X
3.8
is measurable on C. To prove this let
D : s ∈ C | μSs > 0 .
3.9
The measurability of 3.6 on C implies that D ∈ A. Since
qts q0,
s ∈ C \ D,
3.10
it is enough to show that 3.8 is measurable on D. It follows from the definitions of C and D
that the function
s −→
1
μSs
f dμ,
Ss
g dμ ,
s∈D
3.11
Ss
is measurable. Hence, by Lemma 2.3 d, 3.8 is measurable on D.
It is now clear that z is μ-almost measurable on Sx.
Next, we prove that z is μ-integrable over Sx.
To prove this, it is enough to show that the function
s −→ μSsqts,
s∈X
3.12
is bounded on C. Since
μSsqts 0,
s ∈ C \ D,
3.13
we need to verify that 3.12 is bounded on D. By A23 , we can find a t0 > 0 such that
−1
qt < 2
g dμ
·t
∀ t > t0 .
3.14
Sx
It therefore follows from
ts as bs qts,
s∈D
3.15
Journal of Inequalities and Applications
13
that
g dμ 2
μSsts ≤
f dμ μSs
Ss
Ss
−1
g dμ
3.16
ts
Sx
for all s ∈ E : {s ∈ D | ts > t0 }, and hence
⎛
−1 ⎞ 1
⎠≤
μSsts ≤ μSsts⎝1 −
g dμ 2
g dμ
f dμ,
2
Ss
Sx
Ss
s ∈ E.
3.17
Thus
μSsts ≤ 2
f dμ,
s ∈ E,
3.18
Ss
and therefore another application of 3.14 gives
μSsqts
⎧
−1
−1
⎪
⎪
⎪
⎨μSs 2
g dμ
ts ≤
f dμ
g dμ
,
Sx
Ss
Sx
≤
⎪
⎪
⎪
⎩μSxqt
0 ,
⎧
−1
⎪
⎪
⎪
⎨
f dμ
g dμ
, if s ∈ E
Sx
Sx
≤
⎪
⎪
⎪
⎩μSxqt
if s ∈ D \ E,
0 ,
if s ∈ E
if s ∈ D \ E
3.19
and from this the claim follows. Consequently, z is μ-integrable over Sx, as required.
The result is completely proved.
Now we are in a position to prove Theorem 1.5.
Proof. We begin with the proof of c and d.
c To prove that yn ∈ Lloc Dy we use induction on n. Clearly y0 belongs to Lloc Dy .
Let n ∈ N such that the assertion holds. Then Lemma 2.6 yields that yn1 ∈ Lloc Dy . We show
now that the sequence yn is increasing. By our hypotheses on y0 , it follows that y0 ≤ y1 , and
we again complete the proof by induction. Suppose n ∈ N such that yn ≤ yn1 . Then, by
Lemma 2.1 and the induction hypothesis
yn2 x : fx gx
q ◦ yn1 dμ
Sx
q ◦ yn dμ yn1 x,
≥ fx gx
Sx
3.20
x ∈ Dy .
14
Journal of Inequalities and Applications
Since
yn x ≤ yn1 x : fx gx
q ◦ yn dμ,
x ∈ Dy , n ∈ N,
3.21
Sx
Theorem 1.4a implies that
yn x ≤ fx gxμSxqzx,
x ∈ Dy , n ∈ N.
3.22
Because of 3.22 and the fact that yn is increasing, there exists a function y : Dy → 0, ∞
Then y ∈ Lloc Dy is a consequence of yn ∈
such that yn converges pointwise on Dy to y.
Lloc Dy n ∈ N together with 3.22, Lemma 2.1a and, Theorem 1.4b. If x ∈ Dy with
Let
yx
> 0, then according to the continuity of q on 0, ∞, qyn x converges to yx.
0. As we have seen yn x is increasing, and therefore yn x 0 for all
x ∈ Dy with yx
n ∈ N. Consequently
.
q yn x q0 q yx
3.23
Now 1.11 and the monotone convergence theorem give that y is a solution of 1.2.
d We can see exactly as in the proof of c that yn ∈ Lloc Dy , n ∈ N, and the sequence
yn is decreasing. It follows from an easy induction argument that yn is nonnegative for all
n ∈ N. Linking up with the foregoing, there exists a function y : Dy → 0, ∞ such that yn y ∈ Lloc Dy can be shown as in the proof of c.
converges pointwise on Dy to y.
for every x ∈
The continuity of q on 0, ∞ implies that qyn x converges to qyx
Dy .
Assume now that q0 > 0. If x ∈ Dy such that yn x 0 for every large enough n ∈ N,
then
.
q yn x q0 q yx
3.24
If x ∈ Dy such that yn x > 0 for every n ∈ N, then
yn1 x fx gx
q ◦ yn dμ ≥ fx gxq0μSx > 0,
n ∈ N,
3.25
Sx
which leads to yx
> 0. In both cases qyn x converges to qyx
for every x ∈ Dy .
According to 1.11 and the monotone convergence theorem y is a solution of 1.2.
a1 For convergence of the successive approximations
y0 : f,
q ◦ yn dμ,
yn1 x : fx gx
Sx
x ∈ X, n ∈ N
3.26
Journal of Inequalities and Applications
15
to a solution ymin : X → R of 1.2 it suffices, in view of c, to show that f is a solution
of 1.1, which is evident. It remains to prove that if y : Dy → R is a solution of 1.6, then
ymin x ≤ yx for every x ∈ Dy . To this end, it is enough to show that
yn x ≤ yx,
x ∈ Dy , n ∈ N.
3.27
This is true for n 0, since the functions g and q are nonnegative. Let n ∈ N for which the
result holds. Then, because of the nonnegativity of g and the fact that q is increasing,
q ◦ yn dμ
yn1 x : fx gx
≤ fx gx
Sx
3.28
q ◦ y dμ ≤ yx,
x ∈ X, n ∈ N,
Sx
and the proof of the induction step is complete.
a2 Let
Xp : x ∈ X | μSx > 0 .
3.29
Choose x0 ∈ Xp . The set of the upper bounds from LSx0 for the solutions of 1.1 on Sx0 is denoted by U. By Theorem 1.4, U is not empty. Let
U : {u | u ∈ U}.
3.30
Then U is a minorized subset of LSx0 , · , the elements of U are nonnegative. Since
this space is order complete see 12,
3.31
v : inf U
exists. Let v ∈ v, and let y : Dy → R be a solution of 1.1 such that Sx0 ⊂ Dy . yx ≤
ux x ∈ Sx0 , u ∈ U gives that
yx ≤ vx μ a.e. on Sx0 .
3.32
It follows that
yx ≤ fx gx
q ◦ v dμ,
x ∈ Sx0 ,
3.33
Sx
because q is increasing. Since v ∈ LSx0 the proof of Lemma 2.5 shows that q ◦ v ∈
LSx0 , and hence by A4 , the function v1 : Sx0 → R defined by
q ◦ v dμ
v1 x : fx gx
Sx
3.34
16
Journal of Inequalities and Applications
belongs to LSx0 , and therefore v1 ∈ U. It now comes from 3.33 and the definition of v
that
vx ≤ v1 x
3.35
μ a.e. on Sx0 .
Since q is increasing,
q ◦ v1 dμ,
v1 x ≤ fx gx
x ∈ Sx0 ,
3.36
Sx
and thus
v1 x ≤ vx
μ a.e. on Sx0 .
3.37
q ◦ v dμ
3.38
We can see that
vx fx gx
μ a.e. on Sx0 .
Sx
If we set
vx0 : Sx0 −→ R,
q ◦ v dμ,
vx0 x : fx gx
3.39
Sx
then 3.38 shows that vx0 is a solution of 1.2. 3.33 gives that
yx ≤ vx0 x,
x ∈ Sx0 3.40
for every solutions y : Dy → R of 1.1 for which Sx0 ⊂ Dy .
Repeat the preceding construction for every x ∈ Xp . We thus obtain a set of functions
vx ∈ LSx, each a solution of 1.2 on its domain. Moreover, if y : Dy → R is a solution of
1.1, then for every x ∈ Dy ∩ Xp we have
ys ≤ vx s,
s ∈ Sx.
3.41
Introduce the next functions: for every x ∈ X \ Xp with Sx / ∅ let the function vx be defined
on Sx by
vx s : fs.
3.42
Obviously, these functions are also solutions of 1.2 on their domains.
Now let x1 , x2 ∈ X such that Sx1 / ∅ and x1 ∈ Sx2 . Using 3.41, it is easy to verify
that
vx1 s vx2 s,
s ∈ Sx1 ,
3.43
Journal of Inequalities and Applications
17
and therefore Lemma 2.10 is applicable to the solutions vx . This gives a unique solution ymax :
X → R of 1.2 for which
ymax | Sx vx ,
x ∈ X with Sx /
∅.
3.44
It remains to prove that ymax is maximal. Let y : Dy → R be a solution of 1.1, and let
x ∈ Dy . If μSx 0, then
yx ≤ fx ymax x,
3.45
while if μSx > 0, then by 3.41
ys ≤ ymax s,
s ∈ Sx,
3.46
so that Lemma 2.1 implies that
yx ≤ ymax x.
3.47
b We first show that every solution y : Dy → R of 1.2 with Dy /
X has an extension
y1 : Dy1 → R that is a solution of 1.2 such that Dy is a proper subset of Dy1 . This follows
X has
from c as soon as it is realized that every solution y : Dy → R of 1.2 with Dy /
an extension y2 : Dy2 → R that is a solution of 1.1 such that Dy is a proper subset of Dy2 .
Really, in this case the successive approximations determined by y2 converge to a solution
y1 : Dy2 → R of 1.2, which is obviously an extension of y.
That realization can be reached in finitely many steps.
X, and let x0 ∈ X \ Dy .
Let y : Dy → R be a solution of 1.2 with Dy /
i Suppose Sx0 ⊂ X \ Dy .
If Sx0 ∅, then
y1 : Dy ∪ {x0 } −→ R,
y1 :
⎧
⎨yx, if x ∈ Dy ,
⎩fx,
if x x0
⎧
⎨yx,
if x ∈ Dy ,
⎩fx,
if x ∈ Sx0 3.48
is an appropriate solution.
∅, then
If Sx0 /
y2 : Dy ∪ Sx0 −→ R,
y2 :
3.49
is a solution of 1.1 that agrees with y on Dy .
∅ and Sx0 \ Dy /
∅. Let
ii Suppose Sx0 ∩ Dy /
E : x ∈ Sx0 ∩ Dy | μSx > 0 .
3.50
18
Journal of Inequalities and Applications
If E ∅, then the function
y2 : Dy ∪ Sx0 −→ R,
y2 :
⎧
⎨yx,
if x ∈ Dy ,
⎩fx,
if x ∈ Sx0 \ Dy
3.51
is a solution of 1.1 yx fx, x ∈ Sx0 ∩ Dy , thus y2 ∈ Lloc Dy ∪ Sx0 that agrees with
y on Dy .
If E / ∅, then we introduce the functions ux : Sx0 → R, x ∈ E,
ux s :
⎧
⎨ys,
if s ∈ Sx,
⎩fs,
if s ∈ Sx0 \ Sx,
3.52
which all lie in LSx0 , and they are all solutions of 1.1. By Theorem 1.4 b, {ux | x ∈ E}
is a majorized subset of LSx0 , · , . Since this space is order complete see 12,
3.53
u : sup ux
x∈E
exists. Choose u ∈ u. Then u ≥ ux μ-almost everywhere on Sx0 for every x ∈ E, and
therefore Lemma 2.1 yields that
q ◦ u dμ,
ux s ≤ fs gs
s ∈ Sx0 , x ∈ E.
3.54
Ss
Since u ∈ LSx0 the proof of Lemma 2.5 shows that q ◦ u ∈ LSx0 , and hence by A4 ,
the function
s −→ fs gs
q ◦ u dμ,
s ∈ Sx0 3.55
Ss
belongs to LSx0 too. It now follows from u ∈ u and 3.54 that
us ≤ fs gs
q ◦ u dμ,
μ a.e. on Sx0 .
3.56
Ss
We set
u
: Sx0 −→ R,
q ◦ u dμ.
u
s : fs gs
3.57
Ss
By 3.56
u
s ≤ fs gs
q◦u
dμ,
Ss
s ∈ Sx0 .
3.58
Journal of Inequalities and Applications
19
Fix x ∈ E. From Lemma 2.9 with F and G there being {ux | x ∈ E} and {ux | x ∈ E} resp. we
get that us ys μ-almost everywhere on Sx,
and hence
u
s ys,
s ∈ Sx, x ∈ E.
3.59
x fx yx.
Consequently, u
x yx, x ∈ E. If x ∈ Sx0 ∩Dy with μSx 0, then u
We can see that
u
x yx,
x ∈ Sx0 ∩ Dy .
3.60
By what we have already proved that
y2 : Dy ∪ Sx0 −→ R,
y2 :
⎧
⎨yx,
if x ∈ Dy ,
⎩u
x,
if x ∈ Sx0 \ Dy
3.61
is a solution of 1.1 that agrees with y on Dy .
iii Suppose Sx0 / ∅ and Sx0 ⊂ Dy . By an argument entirely similar to that for the
case ii, we can get a solution y2 : Dy ∪ {x0 } → R of 1.1 that agrees with y on Dy .
After these preparations, the proof can be concluded quickly. Let y : Dy → R be a
solution of 1.2, and let P be the set of all solutions of 1.2 which agree with y on Dy . Since
y ∈ P , P is not empty. Partially order P by declaring u1 u2 to mean that the restriction
of u2 to the domain of u1 agrees with u1 . By Hausdorff’s maximality theorem, there exists
a maximal totally ordered subcollection Q of P . Let D be the union of the domains of all
members of Q, and define y : D → R by yx
: ux, where u occurs in Q. It is easy to
check that y is well defined, and it is a solution of 1.2. If D were a proper subset of X, then
the first part of the proof would give a further extension of y,
and this would contradict the
maximality of Q.
e Let y : Dy → R be a solution of 1.2, and let x ∈ Dy .
If μSx 0, then y | Sx f | Sx, so that y is bounded on Sx.
Suppose μSx > 0. Since f and g are bounded on Sx and μSx is finite,
Theorem 1.4 a implies that it is enough to prove the boundedness of the function
s −→ qts,
s ∈ X,
3.62
on Sx. Moreover, we have only to observe that the function
s −→ ts,
s ∈ X,
3.63
is bounded on Sx. To prove this, let
A : s ∈ Sx | μSs > 0 .
3.64
20
Journal of Inequalities and Applications
If fs ≤ c and gs ≤ c, s ∈ Sx, then as ≤ c and bs ≤ c for every s ∈ A, and therefore by
Lemma 2.3d, the function 3.63 is bounded on A. If s ∈ Sx \ A, then the definition of the
function t gives that ts 0. The claim about t is therewith confirmed.
The proof of the theorem is now complete.
Acknowledgment
This work was supported by Hungarian Foundation for Scientific Research Grant no. K73274.
References
1 L. Horváth, “On the associativity of the product of measure spaces,” Acta Mathematica Hungarica, vol.
98, no. 4, pp. 301–310, 2003.
2 J. LaSalle, “Uniqueness theorems and successive approximations,” Annals of Mathematics, vol. 50, pp.
722–730, 1949.
3 I. Bihari, “A generalization of a lemma of Bellman and its application to uniqueness problems of
differential equations,” Acta Mathematica Academiae Scientiarum Hungaricae, vol. 7, pp. 81–94, 1956.
4 D. Baı̆nov and P. Simeonov, Integral Inequalities and Applications, vol. 57 of Mathematics and Its
Applications (East European Series), Kluwer Academic Publishers, Dordrecht, The Netherlands, 1992.
5 D. S. Mitrinović, J. E. Pečarić, and A. M. Fink, Inequalities Involving Functions and Their Integrals
and Derivatives, vol. 53 of Mathematics and Its Applications (East European Series), Kluwer Academic
Publishers, Dordrecht, The Netherlands, 1991.
6 B. G. Pachpatte, Integral and Finite Difference Inequalities and Applications, vol. 205 of North-Holland
Mathematics Studies, Elsevier Science, Amsterdam, The Netherlands, 2006.
7 L. Horváth, “Integral inequalities in measure spaces,” Journal of Mathematical Analysis and Applications,
vol. 231, no. 1, pp. 278–300, 1999.
8 L. Horváth, “Generalizations of special Bihari type integral inequalities,” Mathematical Inequalities &
Applications, vol. 8, no. 3, pp. 441–449, 2005.
9 L. Horváth, “Generalization of a Bihari type integral inequality for abstract Lebesgue integral,” Journal
of Mathematical Inequalities, vol. 2, no. 1, pp. 115–128, 2008.
10 L. Horváth, “Integral equations in measure spaces,” Integral Equations and Operator Theory, vol. 45, no.
2, pp. 155–176, 2003.
11 L. Horváth, “Nonlinear integral equations with increasing operators in measure spaces,” Journal of
Integral Equations and Applications, vol. 17, no. 4, pp. 413–437, 2005.
12 H. H. Schaefer, Banach Lattices and Positive Operators, Die Grundlehren der Mathematischen
Wissenschaften, Band 21, Springer, New York, NY, USA, 1974.
13 E. Hewitt and K. R. Stromberg, Real and Abstract Analysis, vol. 25 of Graduate Texts in Mathematics,
Springer, Berlin, Germany, 1965.
Download